Hardware Support for Advanced Data Management Systems

Reviewer: Floyd S. Shipman

Changes in hardware technology present many new choices to the database machine designer. This paper describes the results of using a sophisticated queueing network model to explore the performance of database machine design. A wide range of architectures and database usage patterns were modeled. For a given trial, the components of the sample architecture were “grown” until the architecture performed adequately for the database load. Then, the resulting architectural configuration and the component costs were used to compute a cost-performance measure. One of the interesting results is that for a large number of active users, the centerlized approach resulted in the least cost. The paper is well written and interesting. More detail about the methodology used can be found in Neches PhD thesis [1].